R/Lrnr_gru_keras.R
Lrnr_gru_keras.RdThis learner supports Recurrent Neural Networks (RNNs) with Gated Recurrent Units (GRU). This learner leverages the same principles as LSTM networks but is more streamlined and thus cheaper to run, at the expense of some loss in representational power. This learner uses the keras package. Note that all preprocessing, such as differencing and seasonal effects for time series, should be addressed before using this learner. Desired lags of the time series should be added as predictors before using the learner.
A learner object inheriting from Lrnr_base with
methods for training and prediction. For a full list of learner
functionality, see the complete documentation of Lrnr_base.
batch_size: How many times should the training data be used to
train the neural network?
units: Positive integer, dimensionality of the output space.
dropout: Float between 0 and 1. Fraction of the input units to
drop.
recurrent_dropout: Float between 0 and 1. Fraction of the units
to drop for the linear transformation of the recurrent state.
activation: Activation function to use. If you pass NULL, no
activation is applied (e.g., "linear" activation: a(x) = x).
recurrent_activation: Activation function to use for the
recurrent step.
recurrent_out: Activation function to use for the output step.
epochs: Number of epochs to train the model.
lr: Learning rate.
layers: How many LSTM layers. Only allows for 1 or 2.
callbacks: List of callbacks, which is a set of functions to
be applied at given stages of the training procedure. Default callback
function callback_early_stopping stops training if the validation
loss does not improve across patience number of epochs.
...: Other parameters passed to keras.
Other Learners:
Custom_chain,
Lrnr_HarmonicReg,
Lrnr_arima,
Lrnr_bartMachine,
Lrnr_base,
Lrnr_bayesglm,
Lrnr_bilstm,
Lrnr_caret,
Lrnr_cv_selector,
Lrnr_cv,
Lrnr_dbarts,
Lrnr_define_interactions,
Lrnr_density_discretize,
Lrnr_density_hse,
Lrnr_density_semiparametric,
Lrnr_earth,
Lrnr_expSmooth,
Lrnr_gam,
Lrnr_ga,
Lrnr_gbm,
Lrnr_glm_fast,
Lrnr_glm_semiparametric,
Lrnr_glmnet,
Lrnr_glmtree,
Lrnr_glm,
Lrnr_grfcate,
Lrnr_grf,
Lrnr_gts,
Lrnr_h2o_grid,
Lrnr_hal9001,
Lrnr_haldensify,
Lrnr_hts,
Lrnr_independent_binomial,
Lrnr_lightgbm,
Lrnr_lstm_keras,
Lrnr_mean,
Lrnr_multiple_ts,
Lrnr_multivariate,
Lrnr_nnet,
Lrnr_nnls,
Lrnr_optim,
Lrnr_pca,
Lrnr_pkg_SuperLearner,
Lrnr_polspline,
Lrnr_pooled_hazards,
Lrnr_randomForest,
Lrnr_ranger,
Lrnr_revere_task,
Lrnr_rpart,
Lrnr_rugarch,
Lrnr_screener_augment,
Lrnr_screener_coefs,
Lrnr_screener_correlation,
Lrnr_screener_importance,
Lrnr_sl,
Lrnr_solnp_density,
Lrnr_solnp,
Lrnr_stratified,
Lrnr_subset_covariates,
Lrnr_svm,
Lrnr_tsDyn,
Lrnr_ts_weights,
Lrnr_xgboost,
Pipeline,
Stack,
define_h2o_X(),
undocumented_learner
if (FALSE) {
library(origami)
data(bsds)
# make folds appropriate for time-series cross-validation
folds <- make_folds(bsds,
fold_fun = folds_rolling_window, window_size = 500,
validation_size = 100, gap = 0, batch = 50
)
# build task by passing in external folds structure
task <- sl3_Task$new(
data = bsds,
folds = folds,
covariates = c(
"weekday", "temp"
),
outcome = "cnt"
)
# create tasks for taining and validation (simplifed example)
train_task <- training(task, fold = task$folds[[1]])
valid_task <- validation(task, fold = task$folds[[1]])
# instantiate learner, then fit and predict (simplifed example)
gru_lrnr <- Lrnr_gru_keras$new(batch_size = 1, epochs = 200)
gru_fit <- gru_lrnr$train(train_task)
gru_preds <- gru_fit$predict(valid_task)
}